*
* Example 14.3, page 572
* Non-linear least squares
*
cal 1970
open data table14-2.prn
data(format=prn,org=columns) 1970:1 1999:1
*
nonlin b1 b2 b3
frml logpop = b1/(1+exp(b2+b3*time))
*
* The following aren't great initial values, and end up needing some extra
* iterations. Note, BTW, that RATS gets different (and better fitting) values
* than cited in the text.
*
compute b1=200.0,b2=b3=0.0
nlls(frml=logpop,iters=500) uspop
*
* Try again from the text coefficients
*
compute b1=1432,b2=1.796,b3=-.0117
nlls(frml=logpop) uspop